The concept
An outlier is unusually distant from the pattern suggested by other observations. It may reflect a recording error, a different process, or a real rare event. Removing it because it spoils a tidy result can distort the evidence. Investigate provenance and context, and make exclusion rules explicit before interpreting the final analysis.
Worked example
A fictional museum records daily visitors as 81, 77, 85, 79, and 840. Staff initially suspect an extra zero. Checking the calendar reveals that the last day included a free festival. The number is valid but belongs to a different operating condition. The team reports ordinary-day attendance separately from festival attendance, explaining the distinction rather than silently deleting the high value.
Limits and judgment
Outlier definitions depend on the distribution and the purpose. A value rare under one model may be common in a heavy-tailed process. Statistical flags identify cases for attention; they do not establish fraud or error. If an exclusion materially changes the conclusion, show results with and without it and explain the substantive reason for the chosen treatment.
FROM UNDERSTANDING TO ACTION
Put it to work.
Write an investigation plan for a delivery record showing a trip ten times longer than usual.
RETRIEVE. REASON. REMEMBER.
Knowledge check.
Choose one answer for each question, then check your reasoning.
Sources & further reading.
Links support the core concepts or identify the book. Examples, exercises, and critical analysis are original Institute material. Source checks are not an independent expert review of every claim.
- NIST: Detection of Outliers ↗
Statistical core concept; all numerical scenarios and interpretations are original instructional examples. Source reviewed 2026-10-19.